동감

To use the code, perform the following steps:
download the files emd.h and emd.c (check the log of changes for latest changes).
In emd.h, modify the line
typedef int feature_t;
to reflect your feature data type. Structures may be used too, for example

typedef struct {
   int X,Y,Z;
} feature_t;
To compute an EMD, call:
float emd(signature_t *Signature1, signature_t *Signature2,
      float (*func)(feature_t *, feature_t *),
      flow_t *Flow, int *FlowSize);
where

Signature1, Signature2:
Pointers to the two signatures that their distance we want to compute.
Dist:
Pointer to the ground distance function. i.e. the function that computes the distance between two features.
Flow:
(Optional) Pointer to a vector of flow_t (defined in emd.h) where the resulting flow will be stored. Flow must have n1+n2-1 elements, where n1 and n2 are the sizes of the two signatures respectively. If NULL, the flow is not returned.
FlowSize:
(Optional) In case Flow is not NULL, FlowSize should point to a integer where the number of flow elements (always less or equal to n1+n2-1) will be written.
Compile emd.c and link it to your code.
The signature data type signature_t is defined in emd.h as:

typedef struct
{
  int n;                /* Number of features in the distribution */
  feature_t *Features;  /* Pointer to the features vector */
  float *Weights;       /* Pointer to the weights of the features */
} signature_t;
The feature data type feature_t is defined in emd.h and should be modified by the user to reflect his feature type.

In special cases, the user might want to change some of the values in the definitions in emd.h:

#define MAX_SIG_SIZE 100
The maximum allowed number of features in a signature
#define MAX_ITERATIONS 100
Maximum number of iterations. For ordinary problems, 100 should be more than enough.
#define INFINITY 1e20
INFINITY should be much larger than any other value in the problem
#define EPSILON 1e-6
EPSILON determines the accuracy of the solution.